Why we built PearPie

What PearPie is

PearPie is an app for chatting with AI without giving up your data. It runs on your computer and your iPhone, and the conversations you have stay on those devices, not on a PearPie server. Your identity is just a key generated on your machine, not a customer record in our database.

If you're curious how the privacy actually works under the hood, the how it works page walks through it. This page is about the why.

The premise

AI assistants are quickly becoming part of how people think, work, and make sense of things. You end up asking an AI the kinds of things you'd usually only discuss with your closest friends or family. That's probably not something you want lingering on a company's server. But the way most AI apps are built, that's exactly where it ends up.

Keeping your data yours shouldn't be a configuration option, an enterprise tier, or something only technical users can pull off. It should be the default. No storage, no training, no profile built on the side: for everyone, not just whoever knew to ask for it.

The clearest case is people whose work is sensitive: finance, medical, legal, anything where pasting into a logged chat box is a regulatory non-starter. They can't really use the current generation of AI tools at all. They're the obvious group, but the principle generalises. No one's conversations should sit on a server, get trained on, or build a profile on the side just because they wanted to ask an LLM something.

PearPie is the version of an AI app that takes that as the starting point.

Built to be simple

The other half of the principle is accessibility. Privacy and ease aren't supposed to be a trade-off. Most privacy-respecting AI today still expects you to be a tinkerer: set up a server, run a command, read a wiki. PearPie doesn't. You install an app, open it, pick a model, talk. The privacy lives in the architecture; the experience is the kind anyone can use.

If something can be private and easy at the same time, both should be the default.

Who it's for

That principle attracts two kinds of users. The product is built for both.

It's a pilot, and that's the deal

PearPie is in its pilot launch. The core experience is there: running models locally, syncing between your devices, accessing premium models on the PearPie Network. But rough edges still exist, some platforms aren't released yet, and the app will keep changing fast. The privacy promises hold from day one, regardless of how early it is.

Where it's headed

Today PearPie can do three things: run open-source models on your own device, let your devices share work directly over peer-to-peer connections, and route requests to premium models on the PearPie Network. Those are top-tier models hosted on European infrastructure. The first two are entirely under your control. The third still relies on cloud providers, even if it's all in Europe and PearPie holds data processing agreements with each of them.

Closing that gap is the long-term direction: more models running locally, a wider peer-to-peer network of trusted devices, and eventually self-hosted European GPU infrastructure for the premium path. The privacy promises above already hold today and will continue to hold; what changes over time is how much of the underlying hardware PearPie controls directly.